MiniMax M3

MiniMax M3 is a non-reasoning model from MiniMax. 48 benchmarks count toward its score, in 7 categories.

availableShows if the model has enough results for an index.
IndexOverall score. 50 is the middle.56.9 ±2.8
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.64/s
Input / 1MUS dollars per 1M input tokens.$0.3
Output / 1MUS dollars per 1M output tokens.$1.2
ContextMaximum tokens in one request.1.05M
EloLMArena rating and rank.1433 (#82)

50 is the middle of the board. The range shows the doubt in the index.

48,540 votes. Elo shows what people prefer. It does not change the score.

CapabilitiesScore per category. 50 is the middle.

50 is the middle
AgenticMulti-step tasks with tools.
60.8
CodingCode writing and repair.
59.4
ReasoningLogic problems and puzzles.
49.4
MultimodalTasks with images and text.
59.2
KnowledgeFacts and expert knowledge.
57.0
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
62.1
MathMath problems.
44.6

Results

48 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result on the index scale.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
Artificial Analysis GPQA DiamondKnowledge92.9%63.8Artificial Analysis
GPQA DiamondKnowledge92.7%63.81 Sept 2026Vals AI
OmniDocBench 1.5Multimodal91.6%OpenAI
τ²-Bench Tool-Agent-User EvaluationAgentic88.9%62.6Victor Barres et al.
Artificial Analysis Harvey LAB-AAAgentic88.4%70.5Artificial Analysis
United States of America Mathematical Olympiad 2026Math85.7%Mathematical Association of America
Video-MME with subtitleMultimodal85.4%Qwen
VideoMMMUMultimodal84.6%Qwen
MMLU ProKnowledge84.2%53.21 Sept 2026Vals AI
BrowseCompAgentic83.5%69.9OpenAI
Artificial Analysis Long Context ReasoningReasoning83.0%65.7Artificial Analysis
Artificial Analysis IFBenchInstruction82.9%73.2Artificial Analysis
LiveCodeBenchCoding82.2%59.01 Sept 2026Vals AI
GPQA diamondKnowledge81.3%53.3none effortEpoch AI
MMMU ProMultimodal81.2%59.41 Sept 2026Vals AI
Software Engineering Benchmark VerifiedCoding80.5%62.1Carlos E. Jimenez et al.
Artificial Analysis MMMU-ProMultimodal78.6%63.1Artificial Analysis
Massive Multi-discipline Multimodal Understanding ProMultimodal78.1%54.4MMMU-Pro authors
LiveBench MathematicsMath76.9%50.625 Jun 2026LiveBench
LiveBench LanguageKnowledge76.8%63.325 Jun 2026LiveBench
LiveBench Data AnalysisReasoning76.2%61.825 Jun 2026LiveBench
BankerToolBenchAgentic76.1%MiniMax
SWE-benchCoding75.0%57.71 Sept 2026Vals AI
GDPval rubricsAgentic74.7%MiniMax
Claw-EvalAgentic74.5%75.5Bowen Ye et al.
LiveBench ReasoningReasoning74.5%59.425 Jun 2026LiveBench
MCP AtlasAgentic74.2%64.0OpenAI
OSWorld-VerifiedAgentic70.1%58.9Tianbao Xie et al.
LiveBench CodingCoding68.2%51.025 Jun 2026LiveBench
SVG-BenchCoding63.7%MiniMax
SWE-bench ProCoding59.0%61.1Xiang Deng et al.
Artificial Analysis Coding IndexCoding58.6%60.2Artificial Analysis
LiveBench Instruction FollowingInstruction57.5%51.025 Jun 2026LiveBench
Terminal-Bench 2.1Agentic53.6%55.721 Sept 2026Vals AI
SkillsBenchCoding51.5%66.6OpenHands11 Sept 2026Vals AI
VIBE V2Coding50.1%MiniMax
OpenHarmony Bench v1.0Coding48.4%57.1OpenHarmony Bench authors
Vibe Code Bench v1.1Coding47.6%62.0OpenHands21 Sept 2026Vals AI
Artificial Analysis SciCodeCoding47.1%58.2Artificial Analysis
Terminal-Bench 2.0Agentic46.1%55.64 Jun 2026Vals AI
OfficeQA ProMultimodal45.1%57.4OfficeQA Pro authors
Terminal-Bench HardAgentic42.4%Artificial Analysis
NL2RepoCoding42.1%58.0MiniMax
LiveBench Agentic CodingAgentic40.7%55.025 Jun 2026LiveBench
Artificial Analysis Humanity's Last ExamKnowledge39.0%66.8Artificial Analysis
GDPval-AA normalizedAgentic36.5%63.6Artificial Analysis
Artificial Analysis EnterpriseOps-GymAgentic32.1%51.4Artificial Analysis
Artificial Analysis Agentic IndexAgentic30.8%62.3Artificial Analysis
Artificial Analysis Intelligence IndexKnowledge29.2%58.9Artificial Analysis
KernelBench HardCoding28.8%MiniMax
OTIS Mock AIME 2024-2025Math26.7%26.5none effortEpoch AI
Code MigrationCoding19.9%57.721 Sept 2026Vals AI
ResearchClawBenchAgentic19.8%InternScience
ProofBench v1.1Math18.0%56.721 Sept 2026Vals AI
Medical Long Context Reasoning (MLCR-AA)Reasoning17.2%52.6Wisedocs and Artificial Analysis
Artificial Analysis Omniscience AccuracyKnowledge16.7%34.5Artificial Analysis
Artificial Analysis AnalystAgentAgentic10.0%48.5Artificial Analysis
Vibe Code Bench 1-100Coding9.2%61.9OpenHands16 Sept 2026Vals AI
Mystery Game PuzzlesReasoning8.0%42.4none effortEpoch AI
OSWorld 2.0Agentic4.6%58.2Mengqi Yuan et al.
Chess PuzzlesReasoning4.0%28.7none effortEpoch AI
Critical Physics TasksReasoning3.7%46.1Artificial Analysis
Agent Arena command recoveryAgentic1.669.215 Sept 2026LMArena
Agent Arena steerabilityAgentic-6.659.915 Sept 2026LMArena
Agent Arena task outcomeAgentic-12.053.815 Sept 2026LMArena

48 benchmarks count, from 54 of 65 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authorsOpenRouter, collected directlyNo licence statedVals AI, collected directlyNo licence stated. Read from the public leaderboard and credited to Vals AIEpoch AI, collected directlyCC BY — free to use and redistribute with attributionLiveBench, collected directlyNo licence stated for the leaderboard. The site repo serving these CSVs has no LICENSE; the harness repo carries upstream Apache-2.0 and MIT copies that cover code, not resultsLMArena, collected directlyCC BY 4.0 (lmarena-ai/leaderboard-dataset on Hugging Face)

Same level, lower price

Qwen3.8-Flash-Next66.4 · Freedots3-note Preview65.5 · FreeApodex 1.1 Mini59.1 · Free

More from MiniMax

MiniMax M2.752.9MiniMax M2.550.9MiniMax M2.150.0MiniMax 01UnrankedMiniMax 2.5Unranked